We investigate deterministic non-preemptive online scheduling with delayed commitment for total completion time minimization on parallel identical machines. In this problem, jobs arrive one-by-one and their processing times are revealed upon arrival. An online algorithm can assign a job to a machine at any time after its arrival. We neither allow preemption nor a restart of the job, that is, once started, the job occupies the assigned machine until its completion. Our objective is the minimization of the sum of the completion times of all jobs. In the more general weighted version of the problem, we multiply the completion time of a job by the individual weight of the job. We apply competitive analysis to evaluate our algorithms. We improve 25-year-old lower bounds for the competitive ratio of this problem by optimizing a simple job pattern. These lower bounds decrease with growing numbers of machines. Based on the job pattern, we develop an online algorithm which is an extension of the delayed-SPT (Shortest Processing Time first) approach to the parallel machine environment. We show that the competitive ratio is at most 1.546 for any even machine number which is a significant improvement over the best previously known competitive ratio of 1.791. For the two-machine environment, this algorithm achieves a tight competitive ratio. This is the first algorithm which optimally solves an online total completion time problem in a parallel machine environment. Finally, we give the first separation between the weighted and unweighted versions of the problem by showing that in the two-machine environment, the competitive ratio of the weighted completion time objective is strictly larger than 1.546.